Exploring the dynamics in experts' participation in a multi-hazards multi-phases emergency management system using textual analysis assisted by large language models

Abstract Experts play important role in the governance system for public emergencies. Different situational and institutional environments across different stages and hazards types in emergency management system shaped complicated interfaces between science and policy, understanding such dynamics is critical for the development of the expert participation regime. This study adopted a large language model (LLM) assisted content analysis method to reveal this dynamics of expert participation in a multi-hazards multi-phases emergency management system in China. There are three main findings: (1) hazards with high epistemic complexity and value consensus foster substantive and sustained expert participation; (2) political sensitivity and decision urgency during crisis tend to constrain the space for the participation of external experts; (3) governance structure, including the organizational types and the openness of procedural design, condition how, when, and to what extent experts knowledge are positioned. By integrating LLM-assisted content analysis methods with an institutional analysis perspective, this study highlights the structural variation in expert participation. It offers new empirical and conceptual insights into the dynamics of knowledge mobilization in emergency management contexts.